Young-Han Kim
We provide a counterexample to Cover's conjecture that the feedback
capacity of an additive Gaussian noise channel under
power constraint
be no greater than the nonfeedback capacity
of
the same channel under power constraint
, i.e.,
.